- changed status to open
- removed comment
MoL should automatically allocate storage for sufficient timelevels of evolved variables
The MoL thorn requires that all evolved variables have at least two timelevels of data, even if the integration method (e.g. RK4) does not need any past timelevels. It uses one of these timelevels as scratch space, potentially in addition to any other scratch space variables. In the case where at least two timelevels are needed for other reasons, this is more efficient than allocating an extra scratch space variable. Rather than requiring the evolution thorn to allocate at least two timelevels of storage for these variables, MoL should ensure sufficient timelevels via the flesh API.
A related issue is that using timelevels for this purpose is wasteful in situations where the timelevel is not used for any other purpose, since the timelevel will be checkpointed even though it is only being used for temporary storage.
Keyword: MoL
Comments (2)
-
reporter -
- removed comment
You may want to have a look at https://bitbucket.org/cactuscode/cactusnumerical/branch/rhaas%2FunlimitedTimelevels which does something similar for the scratch levels.
- Log in to comment